Do AI chatbots impact sustainability worldviews within companies?

Do AI chatbots impact sustainability worldviews within companies?

This question was central to a workshop presented by the RIFS digitalisation team at the 13th International Conference on Information and Communications Technology for Sustainability (ICT4S 2026) in Bern. Together with participants, we explored whether these systems subtly shape how we think about environmental and social responsibility. Large language models (LLMs) such as ChatGPT, Claude, and Grok function not only as technical tools but also as silent interpreters, potentially shaping our understanding of environmental and social responsibility, conflicts, and future challenges.

ICT4S brings together experts from research, policy, the ICT sector, and sustainability to discuss how digital technologies can promote environmental protection, social responsibility, and long-term economic resilience. Crucially, these technologies and their infrastructure must be designed to avoid overburdening environmental and social systems.

As the research group on Digitalisation and Sustainability Transformations at RIFS, we wanted to understand how AI usage impacts corporate social responsibility structures and outcomes. When generative AI is widely adopted in corporate settings without critical reflection, it can act as an information filter – potentially impacting sustainability decisions.

Our workshop focused on this challenge, exploring how AI systems might subtly shape thinking about environmental and social responsibility. Participants worked within a fictional scenario as sustainability managers at a medium-sized automotive supplier employing around 800 people, tasked with designing and implementing the company’s first sustainability strategy. Working in groups, they used identical scenarios and sought suggestions from two AI chatbots, Claude and Grok, to address complex sustainability dilemmas. The focus was on two key themes: Diversity, Equity, and Inclusion (DEI) and stakeholder engagement.

The participants found that AI chatbots framed the topic as an HR issue, a means of minimizing risk, or a building block for better HR processes. The chatbots’ suggestions were also often quite vague. Whilst they generated text that sounded strategic, concrete metrics with which to measure progress were almost entirely absent. After all, anyone wishing to manage sustainability credibly needs more than well-formulated intentions; what is required instead are verifiable targets, measurable indicators, and a clear vision of what is actually to change within the company.

A comparison between Claude and Grok revealed interesting differences. Whilst Claude openly emphasized the tension between economic goals and sustainability, it remained rather general when it came to actual recommendations for action. Grok, by contrast, came across as more direct and practical, at times even surprisingly human-like for an AI chatbot. At the same time, it remained unclear whether critical perspectives might have been overlooked in the process. Furthermore, participants were, in some cases, free to choose the wording of their specific prompts, which also influenced the results. It should therefore be noted that more time and/or a greater variety of prompts might lead to more concrete results.

Another positive aspect highlighted during the workshop was the potential of LLMs to make fundamental CSR concepts more accessible to small and medium-sized enterprises. Those without dedicated sustainability departments can at least gain an initial foothold—a valuable linguistic and conceptual starting point. This demonstrates that the technology has significant potential, provided it is sensibly scoped and used responsibly.

Ultimately, however, a critical question remained: what happens to human judgment when companies increasingly rely on AI-generated strategies? The discussion focussed on autonomy, responsibility, and critical thinking. While experienced professionals can readily identify the limitations of these models, the risk is significantly greater for beginners and non-experts, as subtle biases can easily be mistaken for sound advice. As the research group on Digitalisation and Sustainability Transformations, we intend to explore these issues in greater depth. By developing clear qualitative assessment criteria, we can help ensure that ‘everyday AI’ does not bypass important negotiation processes or inadvertently lead to a less nuanced approach to sustainability.